So, if I'm reading this correctly, it's about licensing then? Because the
options (OLAP, Data Vault) are licensed at the CDB (instance layer) ? That
would make sense if that's the driving factor. Otherwise you would (could)
have different PDBs attached with different licensed components. PDBA
might have DV and PDB2 might have OLAP and you'd have to manage a household
of options for just one CDB. Am I close, or still missing "it"?
Chris
On Tue, Jul 19, 2016 at 11:58 AM, John Mchugh <john.mchugh@xxxxxxxxxx>
wrote:
Hi Chris,
let me see if I can clarify these questions. See inline responses...
On Jul 19, 2016, at 7:49 AM, Chris Taylor <
christopherdtaylor1994@xxxxxxxxx> wrote:
As I'm learning the multi-tenant technology with 12c, I'm struck by the
following ODD choice that Oracle made. I feel certain there's a reason
(and perhaps a good one) why Oracle chose to increase COUPLING instead of
building independent, low dependence modules?
Actually it was not an Oracle decision per se, but Oracle responding to
customer demand.
Here's what I'm talking about: (Note: 2001512.1)
*Oracle recommends the creation of a container database with all options
as that configurationrules out any option related mismatch problems while
unplugging and plugging a PDB from one container to another. However
customers have often asked whether it is possible to create a container
database with a subset of options. Although it is still recommend to create
a CDB with all options, this document outlines a supported method of
creating a CDB with the options that the customer chooses to install.*
And (Note: 1616554.1)
*Oracle generally recommends that a CDB should have all options
installed in order to avoid any issues with the plug-in of a PDB from
another environment.*
It seems to me that this over complicates the situation unnecessarily but
I'm sure there must be a good reason.
Why doesn't it make more sense to have the CDB as a "brain" that doesn't
really care about what options are installed in a PDB. Why does the CDB
need all the options installed to properly plugin another PDB that has some
"x" option installed?
This was exactly the initial intent where the CDB, as the ‘brain’, which
has all of the options available such that any PDB plugged in that requires
a specific option would have that option available. This avoids the
complexities of placement and maximizing consolidation density where if
this were not the case, you would need to map PDBs to CDBs providing the
specific option and consolidation density would be negatively impacted. ULA
customers obviously have no problem w/ this, however, other customers did
not want to incur the licensing implications for all options. Hence, the
MOS note which describes the cumbersome 12.1 steps limit the options
configured in the CDB. This gets much, much better in 12.2 in terms of ease
of configuration and licensing.
Wouldn't it make more sense for the CDB to ambivalent about what options
the PDB has and just be the controller instead of having these created
dependencies between the CDB and the PDBs it controls?
The CDB cannot be ‘ambivalent’ to the options required by the PDB. The
architecture is what we refer to as in-database virtualization where we
have a pristine, master copy of the Oracle data dictionary in CDB$ROOT. For
this reason, CDB$ROOT will have all dictionary references for all options
configured. PDBs use object and/or metadata links to resolve
inter-dictionary dependencies between the PDB dictionary and the CDB$ROOT
dictionary. It is required that the PDB have all or a subset of options
installed in the CDB$ROOT. Otherwise, you get the plug in violation warning
until the dependencies are resolved.
It just seems like this should have raised some engineering red flags
along the way unless there's something I'm not seeing.
Let me know if this answers your questions. This dictionary relationship
is really the core of Multitenant and gives us many opportunities for
future development.
thanks,
jpm
Chris
